[firecrawl-ui] Fix Extract: poll the async job for its result

[obeone]

Summary

Extract reported a generic "Extraction failed" on Firecrawl v2 instances. The cause is in the API adapter: POST /v2/extract answers asynchronously with just { success: true, id } and no status field, but the old polling guard only kept polling when status === 'processing'. With no status on the initial response, it returned the job stub immediately, the view saw no data, and surfaced the failure.

The fix makes waitForExtract poll GET /v2/extract/{id} whenever a job id is present and no inline data came back, stopping on a terminal status (completed, failed, cancelled) or after a bounded number of attempts so a stuck job cannot hang the UI. The endpoint stays /v2/extract on purpose: migrating to /v2/scrape with a json format would drop multi-URL and prompt-only extraction.

Test plan

• npm ci then npm run build succeeds
• npx prettier --check . and npx eslint . pass
• Manual: open Extract, provide a URL plus a prompt or schema, submit, and confirm structured data renders instead of "Extraction failed". (UI click-test still pending: the local browser driver cannot settle on the animated aurora background, so this last step needs a human pass.)